This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] fortran/30207 -- Fix a WHERE and array section problem

Hi Paul,

Thanks for the review/approval.

> I'm sorry that I took a little bit with this - I was checking out some
> simpler alternatives.... which broke things!

I'm a little curious what other approaches you investigated.  After a
little digging, I think I now understand why I thought the front-end
canonicalized "z" and "z(:)" to the same gfc_expr.  It turns out that
it's not the front-end that canonicalizes them, but the scalarizer.

So in trans-array.c:gfc_walk_variable_expr I must have come across the

>>      case AR_FULL:
>>        ...
>>        /* Make sure array is the same as array(:,:), this way
>>           we don't need to special case all the time.  */

gfortran's dependency analysis needs to run prior to (as well as
during) scalarization, so we're given with the raw expressions
instead of the standardized GFC_SS_SECTIONs.

The good news is that I've another clever use of gfc_full_array_ref_p
which I'll post later today, after bootstrap and regression testing.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]